Back to overview Show all

Proceedings (peer-reviewed)

Title of proceedings Proceedings of the 6th Swiss Conference on Data Science
Place Berne
DOI 10.1109/sds.2019.00-10

Open Access

Type of Open Access Repository (Green Open Access)


This study considers a problem of recognizing stress-related conditions in the context of remote stress-monitoring involving wearable biosensors. The participants of the longitudinal study are 18 employees from Public Administration sector wearing biometric devices for around two months in the workplace. The research adopts unsupervised learning approach for exploring stress-related patterns. It investigates physiological signals (Galvanic Skin Response; Heart Rate) combined with additional non-physiological variable (Motion Activity) with help of Gaussian Mixture Model and K-Means classification analysis. Next, it presents bootstrap confidence intervals for evaluating uncertainty of classification. At the same time, it suggests two competing classifiers for identification of stress-related conditions for further research and cross-validation procedure. This work demonstrates that complementing physiological signals with activity-related information improves stress pattern recognition, especially when stress-monitoring is done remotely and user activity is not directly observed during measurements. By implication, this approach may strengthen data governance and enhance the quality of stress-management systems and processes in the workplace.